Classes
struct	ValidationEntry
	one classifier's/regressor's holdout validation outcome (--validate) More...
struct	TopEntry
	one entry of the dupless tracker's --top-k best distinct subsets More...
struct	RegEntry
	one --regularize-margins sweep point: the best subset within margin of the maximum by the secondary criterion More...
struct	TrialEntry
	one --repeat-over-splits trial's search outcome More...
struct	StabilityEntry
	one named stability/similarity measure computed by the stabileval tracker (see the FST4_STAB macro) More...
struct	DafEntry
	one --daf-rank ranked feature More...
struct	PairDafEntry
	one --daf-pairs pairwise (2-feature) DAF statistic More...
struct	SweepEntry
	one --regr-multipliers sweep point (demo11reg) More...
struct	CurvePoint
	one ROC/PR point at a given decision threshold (see extract_curves()) More...
struct	ClassCurve
	one class's ROC and PR curves (see extract_classifier_curves()) More...
struct	CurveEntry
	one curve_over_d point: the best subset of a given size plus its holdout re-validation (the overfitting curve) More...

Public Attributes
std::vector< DIMTYPE >	features
	the final selected subset, sorted ascending
RETURNTYPE	value
	the final subset's criterion value
std::vector< RETURNTYPE >	per_split
std::vector< ValidationEntry >	validation
std::vector< TopEntry >	top_k
std::vector< RegEntry >	regularization
std::vector< TrialEntry >	trials
std::vector< StabilityEntry >	stability
std::vector< DafEntry >	daf
	--daf-rank: DAF ranking from feature-stats
std::vector< ValidationEntry >	daf_validation
std::vector< PairDafEntry >	daf_pairs
std::vector< SweepEntry >	multiplier_sweep
	--regr-multipliers sweep results (demo11reg)
double	best_multiplier
	the sweep's best-by-train-value multiplier
RETURNTYPE	bias
	--bias-estimate result
bool	has_bias
	whether bias was filled (--bias-estimate was given)
std::vector< DIMTYPE >	classified
	--classify-samples predicted classes
std::vector< ClassCurve >	curves
	ROC/PR of the first --validate classifier on the depth-0 holdout.
std::string	curves_classifier
std::vector< CurveEntry >	curve_over_d
	best-per-size search trajectory (criterion axis) + holdout re-validation (validated axis)
std::string	tuned_which
	"svm"\|"rbfsvm" after an optimize protocol; empty = none
double	tuned_c =0.0
double	tuned_gamma =0.0
double	tuned_coef0 =0.0
	the tuned coef0 parameter, meaningful only when tuned_has_coef0
bool	tuned_has_coef0 =false
	whether tuned_coef0 applies (poly/sigmoid LibSVM kernels only)
double	wall_time_s
	wall-clock seconds spent in run_pipeline(); omitted from output under --stable-output
long long	evaluations
	total criterion evaluations (Pipeline::evals); omitted from output under --stable-output

Detailed Description

everything one run_pipeline() call produces: the selected subset plus every optional block gated by its own CLI flag (all empty/default when the flag is not given, so the emitters skip them uniformly across text/json/csv)

Member Data Documentation

◆ curves_classifier

std::string fst4_cli::RunResult::curves_classifier

which classifier produced the curves

◆ daf_pairs

std::vector<PairDafEntry> fst4_cli::RunResult::daf_pairs

--daf-pairs: pairwise DAF statistics (top pairs by pair-DAF0)

◆ daf_validation

std::vector<ValidationEntry> fst4_cli::RunResult::daf_validation

--daf-validate-every: accuracy of incrementally built top-DAF subsets (classifier field holds the subset size)

◆ per_split

std::vector<RETURNTYPE> fst4_cli::RunResult::per_split

filled only when the criterion-protocol splitter yields >1 split

◆ regularization

std::vector<RegEntry> fst4_cli::RunResult::regularization

--regularize-margins sweep results

◆ stability

std::vector<StabilityEntry> fst4_cli::RunResult::stability

stabileval measures over the trials

◆ top_k

std::vector<TopEntry> fst4_cli::RunResult::top_k

the --top-k best distinct tracked subsets (dupless tracker), best first

◆ trials

std::vector<TrialEntry> fst4_cli::RunResult::trials

--repeat-over-splits per-trial results

◆ tuned_c

double fst4_cli::RunResult::tuned_c =0.0

the tuned C parameter

◆ tuned_gamma

double fst4_cli::RunResult::tuned_gamma =0.0

the tuned gamma parameter

◆ validation

std::vector<ValidationEntry> fst4_cli::RunResult::validation

--validate: train on depth-0 train part, test on depth-0 test part; reports accuracy plus the selected --wrapper-metric

The documentation for this struct was generated from the following file:

/Users/psms/dev/devtest1/fst4.cpp

Classes

Public Attributes